Mentioned in This Article
1. CFTC AI trading agents come under the spotlight
On Wednesday, October 28, the US Commodity Futures Trading Commission held a forum on artificial intelligence at which Chairman Mike Selig described the technology as a route to autonomous trading and, in the same breath, listed the regulatory questions it raises. He asked whether developers of AI trading agents must register with the agency and who is liable when such an agent behaves in ways nobody expected.
The event fits a pattern under Selig, whose agency has widened its attention to financial technology, including crypto assets, prediction markets and now AI. The commission regulates derivatives markets in the United States, so any shift in how those contracts are traded falls within its remit.
Selig's tone was optimistic rather than cautionary. He argued that the combination of what he called superintelligence, crypto and modern financial infrastructure could transform markets by widening access to sophisticated tools, improving risk management and creating new kinds of economic activity.
For readers unfamiliar with the agency, the CFTC oversees futures, options and swaps, the contracts through which businesses hedge price risk and traders speculate on everything from interest rates to commodities. Automated trading is not new in those markets, since algorithmic strategies have operated for years. What the chairman described is a further step, in which software does not merely execute a human's instructions but identifies opportunities and acts on them with little or no supervision.
2. Why drop the word artificial?
Panelists at the forum used the term superintelligence, shortened to SI, a phrase favored by President Donald Trump. Their own companies and social media accounts still tend to say artificial intelligence, so the choice of vocabulary was notable.
Selig went further and said the word artificial should be dropped, because in his view it describes the technology inaccurately. The remark signals how the chairman wants the agency to frame the subject: as a genuine form of intelligence to be accommodated by policy, not as an imitation to be treated with suspicion.
He placed the technology in a long historical line, comparing it to the printing press, the steam engine and the internet, each of which reshaped economic life over generations. In his telling, superintelligence will continue a decades-long movement toward ever more autonomous trading systems.
3. What would a trading agent mean for registration and liability?
Selig illustrated the issue with a hypothetical. Suppose a superintelligent agent identifies trading opportunities and executes trades on a user's behalf. The chairman asked whether the developer of that agent would become subject to registration, and who would be responsible if the agent acted in an unforeseen way. His question was direct: does the developer of that agent become subject to registration?
These questions touch long-standing principles of financial regulation. Firms that advise on or execute derivatives trades for others are generally required to register with regulators and meet conduct standards. If software performs part of that role, the law has to decide whether the person who wrote the software, the person who deployed it, or the user who instructed it carries the obligation.
The CFTC already uses AI extensively in its own market surveillance, Selig said, so the agency is both a user of the technology and its overseer. That dual position gives it practical experience, but he nonetheless acknowledged that the growing number of AI trading agents poses questions without settled answers.
The answer will matter for the economics of the whole field. If developers face registration duties, building a trading agent could resemble starting a regulated financial business, with compliance costs and reporting obligations. If liability rests mainly on users, developers may be freer to innovate, yet ordinary investors would bear risks they may not fully understand. Selig's questions leave both outcomes open.
4. Could regulators need a know-your-agent standard?
A second problem concerns identity. Regulators may struggle to determine whom a non-human trading entity actually serves. Financial institutions are required to identify their customers under know-your-customer rules, which are meant to prevent fraud, money laundering and market abuse.
Selig raised the possibility that a new identity standard may be needed, an idea sometimes called know-your-agent, or KYA, to sit alongside existing customer checks. Such a standard would aim to establish which human or organization stands behind an automated trader, so that responsibility for its behavior can be traced.
He did not announce any proposal or rule. The remarks were framed as questions for the industry and the agency to work through, consistent with a forum whose purpose was to gather views.
Know-your-customer rules were built for a world in which an account belongs to a person or a company. An AI agent complicates that model, since one person may run many agents, and a single agent may act for many people. Any identity framework would have to account for those possibilities while avoiding burdens that make legitimate automated trading impractical.
5. How does the crypto experience shape the agency's approach?
Selig said the commission's experience with crypto provides a useful guide. Developers in that sector have long been uncertain whether publishing code, or building an interface, could trigger rules designed for traditional intermediaries such as brokers and exchanges. That uncertainty discouraged some projects and left others operating in a gray area.
The chairman said his leadership has sought to provide clarity on those questions. By drawing the comparison, he suggested that the agency intends to address AI developers with a similar aim, offering definable rules rather than leaving firms to guess.
The parallel is natural, because the two fields share a feature: software that can move value or place orders without a human approving each step. Questions about when a tool becomes a regulated intermediary arise in both.
The comparison also carries a caution. Clarity from a regulator can encourage investment, but it can only be as good as the questions it answers. The questions Selig posed about developer registration and liability are the same kind that crypto builders asked for years, and the industry will be watching to see whether the agency answers them faster for AI than it did for digital assets.
6. The refrigerator that buys its own groceries
To show how far autonomy could go, Selig described a future in which a refrigerator buys groceries on its own, paying with funds it earns by launching a memecoin or by staking ether. The image was meant to be vivid, and it conveyed his point that machines could eventually act as economic participants with their own income and spending.
He argued that regulators must prepare for such a world with care. In his words, the agency has to think long and hard about how to regulate intelligently in this age of superintelligence. The phrasing leaves room for both light-touch and firmer approaches, depending on the risks that emerge.
Memecoins and staking are an interesting choice of examples for a US regulator. They are activities that, only a few years ago, drew heavy scrutiny, and the chairman's use of them as everyday economic acts reflects how far the agency's stance toward crypto has moved.
7. What did Brett Harrison say about AI and trading?
Among the panelists was Brett Harrison, chief executive of Architect, a firm building an AI derivatives exchange, and a former president of FTX US, the crypto exchange that collapsed in a global fraud scandal. Harrison said consumer AI tools such as Claude, ChatGPT and Gemini allow users to turn unstructured information, for instance news items, into trading decisions.
He said the technology is letting large firms expand their analysis exponentially while lowering the barriers to entry for smaller firms. That combination suggests that competitive advantages once held by well-resourced institutions may narrow.
Harrison also said that coding tools built on superintelligence let market participants, from retail traders to institutions, create algorithms that access markets, ingest data and generate fair-value signals. Many of them, he said, now use black box algorithms to trade derivatives. A black box algorithm is one whose internal logic is not easily inspected, which is precisely why questions of accountability and oversight become harder.
Harrison's remarks describe a market in which analysis that once required teams of researchers can be done by a single person using widely available tools. That could improve access, as Selig suggested, but it also means that more trading decisions may rest on automated interpretations of news, a point that connects directly to the chairman's concerns about accountability.
8. What happens next at the CFTC?
The commission is holding roundtables on emerging issues and has announced a further meeting on 24-hour trading, scheduled for December 15. Round-the-clock markets would interact closely with automated agents, which do not need to sleep and can trade at any hour.
The commission's sister regulator, the Securities and Exchange Commission, held a similar roundtable in September 2026, so both agencies are gathering input on how new technology fits within existing law.
No rule changes were announced at the forum. For now, the open questions Selig posed, covering registration of developers, liability for unexpected behavior and the possible need for agent identity checks, remain items for discussion. Market participants building AI trading tools have a window in which to put their views before the agency, and the December meeting is the next scheduled opportunity.
Taken together, the forum showed an agency that welcomes the technology while recognizing that its rulebook was written for human actors. The chairman's questions on registration, liability and identity define the agenda for the months ahead, and the outcome will help determine how automated traders are allowed to operate in US derivatives markets.